When artificial intelligence and death meet, a flood of headlines is rarely far behind. Last month, they were about a self-driving Uber that ran over and killed a woman as she crossed a road in Arizona. And then a Tesla, driven by its software, hit a central reservation in the US, killing the driver.
While there is , there is one area where there is broad support for keeping AI at bay: weapons that can target and fire without human oversight.
Cue more headlines about AI and death this week with the announcement of a of a South Korean university over fears it would work on such weapons. More than 50 AI and robotics experts said they would stop collaborating with tech research institute if this were the case.
Advertisement
In February, it was establishing a Research Center for the Convergence of National Defense and Artificial Intelligence, alongside sponsor Hanwha Systems, an arms-maker. However, the centre鈥檚 stated aims didn鈥檛 exclude weapons that let machine intelligences choose and destroy their own targets.
The signatories demanded that the institution doesn鈥檛 鈥渄evelop autonomous weapons lacking meaningful human control鈥�. In response, KAIST last night affirmed that it wouldn鈥檛 create such weapons and that the research centre was committed to meaningful human control in its projects. It looked likely that the boycott would be called off.
, professor of AI at the University of New South Wales, Australia, and lead signatory of the letter, broadly welcomed KAIST鈥檚 assurance. But he still worries about the unpredictability of military AI systems, especially those that use machine learning.
Ban ahead?
The protest comes amid a UN move to quash autonomous, target-destroying weapons. At a meeting in Switzerland , the UN will further discuss the issue, with 22 nations already seeking to ban such weapons.
This isn鈥檛 about artificial intelligence having no place in the military. That was made clear in to an upcoming report from the UK鈥檚 House of Lords on AI鈥檚 prospects, including its role in weaponry. In giving evidence, , a roboticist at the University of Sheffield, UK, and another signatory of the KAIST letter, made clear that his opposition is to weapons that use autonomy for 鈥渢arget selection and the application of violent force鈥�.
His thoughts were echoed by other expert witnesses: AI should be free to predict battlefield risk, locate improvised explosive devices, disable bombs and simulate disaster response. To block any of this would deny armed forces the very real benefits of some promising, life-saving technologies.
But in a world where unpredictable technology such as machine learning is embraced by the military, expect more protests.
Read more: Killer robots: It鈥檚 time to decide who pulls the trigger; Letting robots kill without human supervision could save lives
